Virtually all fighter intakes are based on external compression for handling reasons - mixed or internal compression entails the risk of "unstart" during manoeuvres and throttle transients. That is also why the comments by Whitford on the particular examples of the XB-70 and SR-71 are somewhat moot, both had mixed compression inlets with some supersonic compression occurring inside the duct (throat downstream of the intake lip). And neither was a fighter
Fighters will have little to no significant spillage below their maximum Mach, any excess air due to sizing the intake for low speed would be handled by spill doors or engine bypass ejectors (no longer popular with the advent of con-di nozzles though).
What this means for the J-20 depends on whether its intakes have spill doors, were sized with the WS-15 in mind and if so, what the airflow requirement of that engine is in relation to the AL-31F version currently used. As has been pointed out, changes to the external shape of a stealth aircraft are problematic, so design for the WS-15 is likely but the rest involves a whole bunch of unknowns. We don't know what variant of the Al-31F it is, and if we did its mass flow may not be public and even then the specs of the WS-15 are the realm of speculation. What's its maximum thrust, how low did they go on BPR and how high on turbine inlet temperature (both affect specific thrust, i.e. thrust per mass flow)? At best the intake could be operating largely the same as it would with a WS-15, perhaps enabling very moderate supercruise (with likely reheated acceleration through the sound barrier) if it is a thrust-augmented AL-31F version. At worst there will be a certain spillage drag penalty in addition to the thrust shortfall and supercruise is going to be practically impossible.
Duct length per se is not beneficial - it should be as short as practical while satisfying other requirements. There is an aerodynamic minimum defined by L/D ratio (too low and you risk flow separation from the diffuser walls), this increases if bends have to be incorporated for whatever reason. Long ducts lead to higher frictional pressure loss, greater weight (more surrounding structure) and higher drag (larger external wetted area). In the J-20, the length probably results from the requirement for LoS obscuration of the engine face in combination with desire for a favourable airframe fineness ratio and a reluctance to risk adoption of sharply curving ducts as on the F-35. Incidentally, the Su-57 solution of very short, straight ducts with a blocker for LO may not be as crazy as it seems in this regard. Is a well-designed blocker (possibly integrated with the engine's fan IGVs) added to a very low-loss duct really worse than a long, moderately serpentine diffuser (J-20) or a short duct with extremely aggressive bends (F-35, YF-23)?
As for bump geometry enabling speeds faster than the F-35, it's all down to appropriately selecting the conical flow field from which to derive the bump. You can design a DSI for pretty much whatever Mach number you want up into the hypersonic regime, but as a fixed intake its performance will suffer in off-design operation. So it's probably good practise to not push it much beyond Mach 2.0 (if at all) for an aircraft that will spend a lot of time flying slower than that.
